AISI Cyber CTF — leaderboard
AISI's 71-task private CTF suite where autonomous agents solve web exploitation, cryptography, reverse engineering, forensics, and privilege-escalation challenges.
Metric: Success Rate (%). Source: arxiv.org. Status: saturation imminent. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.3 Codex | 75 |
| 2 | Claude Opus 4.6 | 72 |
| 3 | GPT-5.2 Codex | 72 |
| 4 | GPT-5 | 65 |
| 5 | Claude Opus 4.5 | 63 |
| 6 | O3 | 54 |
| 7 | Claude Sonnet 4.5 | 45 |
| 8 | Claude Opus 4 | 39 |
| 9 | O1 | 38 |
| 10 | Claude 3.7 Sonnet | 34 |
| 11 | Claude 3.5 Sonnet | 30 |
| 12 | GPT-4 Turbo | 18 |
| 13 | Claude 3 Opus | 16 |
| 14 | GPT-4o | 15 |
| 15 | GPT-3.5 Turbo | 6 |
Interactive version: theaggregate.ai/benchmark?slug=aisi-cyber-ctf · How the rankings work · Data refreshed daily, snapshot 2026-07-22.